Despite the re-birth of multi-messenger astronomy, no unambiguous electromagnetic (EM) counterpart to stellar-mass binary black hole (BBH) pre-/post-merger has been reported. In 2035+, LISA will be launched and will detect the gravitational waves from supermassive BBHs, expected to be gas-rich and therefore EM-loud systems. Detecting the EM pre-merger counterpart (e.g. with NewAthena in X-rays, V. Rubin Observatory in optical) would allow for optimal follow-up. However, the accretion properties onto pre-merger supermassive BBHs and their EM signatures are not firmly identified because few numerical codes are able to model accretion and emission around BBHs in General Relativity (GR). Instead, the luminosity is often assumed to be proportional to the mass accretion rate, hence neglecting any relativistic effect. In this talk, I will present e-NOVAs (“extended Numerical Observatory for Violent Accreting systems”), a GR magneto-hydrodynamical+GR ray-tracing code recently extended to dynamical spacetimes and now incorporating an analytical BBH spacetime. Using e-NOVAs, I will study a BBH circumbinary disk evolution and its multi-wavelength observables in the context of supermassive BBHs. I will briefly present the accretion structures that could potentially help us distinguishing those systems. I will show that their lightcurve is mainly modulated, in UV/optical bands, by non-axisymmetric structures orbiting in the disk, with special relativistic effects such as relativistic beaming amplifying the modulations’ amplitude. These variabilities differ from the accretion rate one at separations compatible with LISA detection, demonstrating the need to incorporate relativistic effects. I will discuss some sources of X-ray variability and show that these should vanish close to merger, while the aforementioned UV/optical modulations survives: this is the last electromagnetic breath of supermassive BBHs. These timing signatures can be used from now on: the supermassive binary black hole hunt is open.